Transit method is used to detect the presence of exoplanets around distant stars. Here's a helpful visualization:
May 1, 2019
Introduction
Data and Process
Model and Learning
Results
Color:
- Red: Detected anomaly
Scope for Improvement
- Hyperparameters such as sequence length, number of hidden units in LSTM layers, number of dense layers, number of neurons in dense layer, etc. can be tuned.
- Bi-directional LSTM autoencoder can be used instead of directional LSTM autoencoder.
- Detections above mean can be explicitly removed before passing to phase folding. This will remove all anomalous events of increase in brightness of star, which will leave out only anomalous events of decrease in brightness of star caused by eclipsing binary star or exoplanet or comet/asteroid/debris